tiktok-ad-research
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a bash helper script (scripts/crawlora.sh) to perform API requests. This script demonstrates high security standards: it validates the character set of the CRAWLORA_API_KEY environment variable to prevent injection; it enforces a strict whitelist of 10 allowed TikTok API paths; it uses a temporary configuration file with 600 permissions to pass credentials to curl (avoiding process-list exposure); and it correctly uses URL-encoding for query parameters while restricting the method to GET.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external TikTok ad data, creating a potential surface for indirect prompt injection. 1. Ingestion points: Ad metadata (titles, descriptions, performance metrics) fetched from api.crawlora.net. 2. Boundary markers: The skill does not define specific delimiters for untrusted ad content. 3. Capability inventory: The agent is restricted to making GET requests to specified vendor API endpoints via the provided script. 4. Sanitization: The script sanitizes query parameters, but the agent processes the retrieved ad text without specific sanitization.
- [SAFE]: All external communication is directed to the vendor's verified API domain (api.crawlora.net), matching the skill's author context.
Audit Metadata